Text complexity in elementary classrooms is typically measured by traditional readability tools, which rely on surface-level measures of word and sentence complexity. Theoretical and empirical work on text complexity, however, indicates that additional measures of semantics, syntax, and discourse structure may be equally important for understanding how young children process texts. This study evaluated the components of two available text analysis tools representing different generations of readability: the Lexile Framework and the Coh-Metrix Text Easability Assessor. Using multilevel modeling, informational passage readings were nested in 5133 second-grade children in 459 schools to identify text variables that predicted changes in oral reading rate. Coh-Metrix dimensions of word concreteness, referential cohesion, and deep cohesion were uniquely associated with oral reading rate after controlling for traditional readability components. A full model including both the Lexile components and three Coh-Metrix dimensions provided the best fit to the data. Results suggest that measures of semantic complexity and text cohesion are needed to improve prediction of text difficulty for young readers and better inform text selection in the early grades. Implications for policy and practice under the Common Core State Standards, which emphasize reading complex texts across the gradespan, are discussed.